Impact of the Input Representation on Pulmonary Hypertension Detection from Heart Sounds through CNNs

Pulmonary hypertension (PH) is a hemodynamic condition describing elevated pulmonary artery pressure. To date, right heart catheterism is the gold standard<br/>diagnostic test for PH, but it is an invasive and expensive procedure. Deep learning (DL) techniques applied to heart sounds have previously shown promising performances for PH screening. In this work, we analyze<br/>the impact of different input representations for PH detection with convolutional neural networks (CNNs). We found that considering each heartbeat as an independent input yielded systematically lower performance than considering the recordings as a whole: preserving the information about the variability over the heartbeats is key. Time-domain feature maps outperformed handcrafted features and combining the time- and frequency-domain proved consistently most effective. Reducing the number of heartbeats to 30 did not affect<br/>the performance, and even reducing to 10 beats preserves the diagnostic value. The proposed analysis moves one step further the applicability of DL for PH detection from heart sounds in the clinical practice.<br/>